XPLR: Network Virtualization Using Dynamic FPGA Reconfiguration

利用FPGA动态重构实现网络虚拟化随着Internet的发展和演进，越来越多的网络应用将被部署以适应业务和社会需求。这些日益多样化的网络应用无疑将加剧对网络服务频谱的需求。通常，网络应用程序在安全性、可预测性和吞吐量方面需要截然不同的性能需求。虽然可以构建物理上独立的网络来满足这些不同的性能限制，但在许多情况下，需要一个共同的物理基板来最大限度地减少设备投资、运行成本和功耗。最近介绍了几种共享网络节点和链路的虚拟网络实现。为了允许快速分配系统资源和灵活的编程环境，这些系统被部署在通用处理器中，每个虚拟网络所需的物理资源由操作系统动态分配。尽管这种方法已被证明是可行的，但通用处理器的串行特性限制了虚拟网络的性能。在这个项目中，将开发一种新的基于硬件的虚拟网络构建方法，提供可扩展性和可编程性。计划中的计算平台使用可重新配置的现场可编程门阵列（FPGA）来实现一个或多个独立的唯一路由器，这些路由器已被定制以满足特定的虚拟网络需求。随着虚拟网络数量及其特性的变化，FPGA中的硬件可以重新配置以支持更新的要求。所部署的虚拟路由器的管理由在附带的微处理器上执行的资源管理器执行。为了评估该方法，将创建一系列软件工具和硬件模块。智力优势：该项目代表了为创建和部署可重构虚拟网络而开发集成环境的积极努力。这种协同工作利用了FPGA系统、FPGA资源管理和网络路由器开发方面的先进技术，允许创建可扩展的高性能虚拟网络。将开发一个虚拟网络架构层来处理动态变化的网络需求。架构资源将由一种新的资源分配算法进行管理，该算法将允许在虚拟网络性能限制下有效利用可用的FPGA区域。该算法将定期运行，以允许FPGA资源的动态再平衡。系统的所有硬件和软件组件将被设计为允许不熟悉FPGA设计的用户有效地使用系统。系统的实时性能将在pi中进行评估。马萨诸塞大学阿默斯特分校的网络实验室在实际工作负载下进行了研究。更广泛的影响：由于网络的普遍性和解决方案的可伸缩性，这项工作可能产生更广泛的影响。研究人员计划开发两个具体项目来扩大这项工作的影响，包括：1)虚拟网络的新本科课程，重点关注可扩展的现实世界系统；2)新技术在美国国家科学基金会资助的协同与自适应大气传感工程研究中心（CASA）的应用，该中心已经有了一个精心设计的基础设施，可以影响波多黎各大学的气象学家、应急管理人员和代表性不足的群体。

Network Virtualization Using Dynamic FPGA ReconfigurationAs the Internet grows and evolves, increasingly diverse network applications will be deployed to accommodate business and social needs. These increasingly diverse network applications undoubtedly will exacerbate the demand for a spectrum of network services. Often, network applications call for strikingly divergent performance requirements in terms of security, predictability, and throughput. Although physically separate networks could be constructed to meet these varied performance constraints, in many cases, a common physical substrate is needed to minimize equipment investment, operating cost, and power consumption. Several virtual network implementations which share network nodes and links have recently been introduced. To allow for the rapid allocation of system resources and a flexible programming environment, these systems are deployed in general-purpose processors and the physical resources required by each virtual network are dynamically allocated by an operating system. Although this approach has been shown to be feasible, the serial nature of general-purpose processors limits virtual network performance. In this project, a new hardware-based approach to virtual network construction that provides scalability and programmability will be developed. The planned computing platform uses a reconfigurable field-programmable gate array (FPGA) to implement one or more individual unique routers that have been customized to specific virtual network needs. As the number of virtual networks and their characteristics change, the hardware in the FPGA can be reconfigured to support the updated requirements. The management of the deployed virtual routers is performed by a resource manager which is executed on an accompanying microprocessor. To evaluate the approach, a series of software tools and hardware modules will be created.Intellectual merit: This project represents an aggressive effort to develop an integrated environment for the creation and deployment of reconfigurable virtual networks. This coordinated effort takes advantage of advances in FPGA systems, FPGA resource management, and network router development to allow for the creation of scalable, high performance virtual networks. A virtual network architectural layer will be developed to address dynamically changing network requirements. Architectural resources will be managed by a new resource allocation algorithm that will allow for the effective use of the available FPGA area under virtual network performance constraints. This algorithm will run periodically to allow for dynamic rebalancing of FPGA resources. All hardware and software components of the system will be designed to allow for effective system use by users that are unfamiliar with FPGA design. The real-time performance of the system will be evaluated in the PIs? networking laboratory at the University of Massachusetts, Amherst under realistic workloads.Broader impact: The potential for broader impact from this work is substantial due to the ubiquity of networking and the scalability of the solution. The researchers plan to develop two specific programs to broaden the impact of the work including: 1) a new undergraduate curriculum in virtual networking, focused on scalable real-world systems, and 2) application of the new technologies in the NSF-sponsored Engineering Research Center for Collaborative and Adaptive Sensing of the Atmosphere (CASA) which already has an elaborate infrastructure in place to impact meteorologists, emergency managers, and under-represented groups at the University of Puerto Rico.